Motivation

test hypothesis
- test and idea or model, find out how things work
search for facts
- study causal effects without having a model in mind first
whisper in the ears of princess
- make suggestions for changes based on data
- test small & cheap interventions to decide which one to scale
get right what others got wrong
- disprove a wrong/weak work
- gather further evidence for an insight

Rules

no deception
- unless you are in marketing
real financial incentives, whenever possible
- sometimes impossible, too expensive, not ethically OK

control: the change between 2 groups
- e.g. heels vs no heels
factor levels:
- 2 levels: heels vs no heels
- 3 levels: <20, 20<22, >22
- …
- n levels
full factorial design
- testing all combinations of factor levels
  - no theoretical limitation, but a practical limitation
between-subject vs within-subject design
- between: each treatment on different subjects
- within: each subject gets different treatments
  - cheaper
  - have a look at Repetition

possibility to study Interaction Effects
- how do different independent variables interact with one another
  - are heels more effective in summer than in winter?
cost/feasibility
- cannot add too many independent variables